learning_rate: 0.0002
train_batch_size: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 32
optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.03
num_epochs: 3

Training Loss	Epoch	Step	Validation Loss	Json Validity	Structure Similarity	Value Accuracy	Field F1	Value Similarity	Edit Distance	Overall Score
11.4229	0.2	5	0.8701	1.0	0.0227	0.0	0.0406	0.0	0.9300	0.1966
5.6784	0.4	10	0.4394	1.0	0.1876	0.1139	0.2742	0.2194	0.4427	0.2448
3.6699	0.6	15	0.3396	1.0	0.4878	0.5354	0.6166	0.6651	0.3013	0.5237
3.099	0.8	20	0.2961	1.0	0.5323	0.5506	0.6440	0.7597	0.2614	0.5554
2.7428	1.0	25	0.2745	1.0	0.4704	0.5	0.5837	0.6403	0.3187	0.5055
2.655	1.2	30	0.2562	1.0	0.4981	0.5862	0.6129	0.6817	0.2968	0.5386
2.2888	1.4	35	0.2435	1.0	0.6938	0.6108	0.7794	0.8188	0.1663	0.6158
2.2156	1.6	40	0.2369	1.0	0.5807	0.5612	0.6767	0.7442	0.2322	0.5624
2.2296	1.8	45	0.2295	1.0	0.7989	0.6980	0.8646	0.8903	0.1325	0.6782
1.9707	2.0	50	0.2248	1.0	0.5920	0.7050	0.6949	0.8421	0.2181	0.6178
2.0616	2.2	55	0.2202	1.0	0.7956	0.6908	0.8518	0.8293	0.1212	0.6566
1.6847	2.4	60	0.2196	1.0	0.7868	0.7429	0.8512	0.8743	0.1115	0.6745
1.8298	2.6	65	0.2177	1.0	0.7496	0.5988	0.8202	0.8110	0.1443	0.6243
1.7607	2.8	70	0.2165	1.0	0.7169	0.8032	0.7995	0.9052	0.1544	0.6811
1.9593	3.0	75	0.2165	1.0	0.7462	0.7108	0.8143	0.8335	0.1331	0.6485

veronikayu/paligemma-imgtojson-0013

paligemma-imgtojson-0013